Re: NS geo-distribution
On 04/30/13 03:30, Dave Warren wrote: On 2013-04-30 00:49, Sten Carlsen wrote: Don't forget that most users will get the address out of "some" cache, not directly from the authoritative servers. Absolutely. This is even more true in our case as many of our clients are serve very local areas and 2-3 ISPs and 3-4 mobile providers probably cover 90%+ of their clients. On 2013-04-29 21:48, Chris Buxton wrote: RTT means almost always hitting the fastest server. My concern with relying on RTT is that since most of our sites are very low volume, will it be effective or does it work better when a host has higher traffic? How long do resolvers remember a particular NS's RTT? We have a handful of Europe based clients, but their number is quite small, so I'm not sure if we'd be significantly hurting the majority by introducing a high-latency server into the mix or not, or even how to evaluate the results. I realize I've probably spent more time thinking about it than I'll possibly save anyone else anyway, so perhaps that's my answer. I appreciate all the input. -- Dave Warren http://www.hireahit.com/ http://ca.linkedin.com/in/davejwarren I may be late to the party, but I am just finishing a project to move LCR's tertiary name server. Over the years, I have been amazed at how small and quick DNS traffic is. With caching, it gets even quicker as far as the client is concerned. Even with a few Europe based clients, dns won't be the slow part. And after reading some material on dns diversity, I decided to move one of my name servers to a hosting company. I picked a low priced company with a virtual machine running OpenSuSE. Installed NAMED plus the RRL patches from source and I was done. Picked up one IPv4 address and one IPv6 address and starting configuring zones. The virtual runs nothing else and I am paying $20/month(should convert to yearly billing and save a couple more bucks). I am in the Chicago area and the virtual machine is in Dallas, TX. Sounds like excellent geo diversity to me. Plus Texas has it's own power grid. So we even have major power grid separation. The dns diversity article I found stated that it's better to get the right IP address and not be able to get to that IP address than to not be able to find any DNS servers for the zone. Email is handled more politely in that respect especially. Plus I remember that the Internet is a best effort network. There is no guarantied connectivity on the Internet. Lyle Giese LCR Computer Services, Inc. P.S. Maybe you would like to use that box I have for a tertiary server. It's got plenty of cpu cycles and extra bandwidth under that hosting package! ___ Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe from this list bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: NS geo-distribution
Dave Warren wrote: > > With the vast majority of our customers being in North America (probably 75% > of users are in Canada), would it make sense to add a Europe based NS or would > this tend to return slower results on average since a potential user would > have a 1/3 chance of hitting a NS with a higher latency? If you can get an anycast secondary then it can be in North America and Europe (and Asia and Africa) at the same time :-) Tony. -- f.anthony.n.finchhttp://dotat.at/ Forties, Cromarty: East, veering southeast, 4 or 5, occasionally 6 at first. Rough, becoming slight or moderate. Showers, rain at first. Moderate or good, occasionally poor at first. ___ Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe from this list bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: NS geo-distribution
On 30/04/13 10:30, Dave Warren wrote: > On 2013-04-30 00:49, Sten Carlsen wrote: >> Don't forget that most users will get the address out of "some" >> cache, not directly from the authoritative servers. > > Absolutely. This is even more true in our case as many of our clients > are serve very local areas and 2-3 ISPs and 3-4 mobile providers > probably cover 90%+ of their clients. Which essentially means that the time to get data from the authoritative servers is irrelevant in almost all cases. Availability vs. network problems would be more important in my book. > > > On 2013-04-29 21:48, Chris Buxton wrote: >> RTT means almost always hitting the fastest server. > > My concern with relying on RTT is that since most of our sites are > very low volume, will it be effective or does it work better when a > host has higher traffic? How long do resolvers remember a particular > NS's RTT? > > We have a handful of Europe based clients, but their number is quite > small, so I'm not sure if we'd be significantly hurting the majority > by introducing a high-latency server into the mix or not, or even how > to evaluate the results. > > I realize I've probably spent more time thinking about it than I'll > possibly save anyone else anyway, so perhaps that's my answer. > > I appreciate all the input. > > -- > Dave Warren > http://www.hireahit.com/ > http://ca.linkedin.com/in/davejwarren > > > > ___ > Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe > from this list > > bind-users mailing list > bind-users@lists.isc.org > https://lists.isc.org/mailman/listinfo/bind-users -- Best regards Sten Carlsen No improvements come from shouting: "MALE BOVINE MANURE!!!" ___ Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe from this list bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: NS geo-distribution
On 2013-04-30 00:49, Sten Carlsen wrote: Don't forget that most users will get the address out of "some" cache, not directly from the authoritative servers. Absolutely. This is even more true in our case as many of our clients are serve very local areas and 2-3 ISPs and 3-4 mobile providers probably cover 90%+ of their clients. On 2013-04-29 21:48, Chris Buxton wrote: RTT means almost always hitting the fastest server. My concern with relying on RTT is that since most of our sites are very low volume, will it be effective or does it work better when a host has higher traffic? How long do resolvers remember a particular NS's RTT? We have a handful of Europe based clients, but their number is quite small, so I'm not sure if we'd be significantly hurting the majority by introducing a high-latency server into the mix or not, or even how to evaluate the results. I realize I've probably spent more time thinking about it than I'll possibly save anyone else anyway, so perhaps that's my answer. I appreciate all the input. -- Dave Warren http://www.hireahit.com/ http://ca.linkedin.com/in/davejwarren ___ Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe from this list bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: NS geo-distribution
Don't forget that most users will get the address out of "some" cache, not directly from the authoritative servers. On 30/04/13 6:48, Chris Buxton wrote: > On Apr 29, 2013, at 9:01 PM, Dave Warren wrote: >> With the vast majority of our customers being in North America (probably 75% >> of users are in Canada), would it make sense to add a Europe based NS or >> would this tend to return slower results on average since a potential user >> would have a 1/3 chance of hitting a NS with a higher latency? > RTT means almost always hitting the fastest server. > > Chris Buxton > ___ > Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe > from this list > > bind-users mailing list > bind-users@lists.isc.org > https://lists.isc.org/mailman/listinfo/bind-users -- Best regards Sten Carlsen No improvements come from shouting: "MALE BOVINE MANURE!!!" ___ Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe from this list bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: NS geo-distribution
On Apr 29, 2013, at 9:01 PM, Dave Warren wrote: > With the vast majority of our customers being in North America (probably 75% > of users are in Canada), would it make sense to add a Europe based NS or > would this tend to return slower results on average since a potential user > would have a 1/3 chance of hitting a NS with a higher latency? RTT means almost always hitting the fastest server. Chris Buxton ___ Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe from this list bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: NS geo-distribution
On 2013-04-29 21:35, Gary L. Burnore wrote: I would contend that fast inititial page load times is achieved through blazing web servers and a wide data path. It sure doesn't hurt, but introducing ~200ms of DNS lookups sure won't make things any faster. -- Dave Warren http://www.hireahit.com/ http://ca.linkedin.com/in/davejwarren ___ Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe from this list bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
RE: NS geo-distribution
> -Original Message- > From: bind-users-bounces+gburnore=netbasix@lists.isc.org > [mailto:bind-users-bounces+gburnore=netbasix@lists.isc.org] On > Behalf Of Dave Warren > Sent: Tuesday, April 30, 2013 12:02 AM > To: Bind Users Mailing List > Subject: NS geo-distribution > > Thoughts about how to best distribute NS servers hosting authoritative > data for our zones? We currently host only about 300 zones so all zones > live on all 3 servers but we're looking at replacing 1-2 of our NS > servers due to reliability of the current servers. > > My thinking is to just pick three sites in Canada/US with good > connectivity and host there, but I'm getting some pressure to pick a > location in EU as well. > > With the vast majority of our customers being in North America > (probably 75% of users are in Canada), would it make sense to add a > Europe based NS or would this tend to return slower results on average > since a potential user would have a 1/3 chance of hitting a NS with a > higher latency? > > I realize that the difference isn't very significant in the grand > scheme of things, but it's always nice to shave a few ms off of initial > page load times. I would contend that fast inititial page load times is achieved through blazing web servers and a wide data path. ___ Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe from this list bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
NS geo-distribution
Thoughts about how to best distribute NS servers hosting authoritative data for our zones? We currently host only about 300 zones so all zones live on all 3 servers but we're looking at replacing 1-2 of our NS servers due to reliability of the current servers. My thinking is to just pick three sites in Canada/US with good connectivity and host there, but I'm getting some pressure to pick a location in EU as well. With the vast majority of our customers being in North America (probably 75% of users are in Canada), would it make sense to add a Europe based NS or would this tend to return slower results on average since a potential user would have a 1/3 chance of hitting a NS with a higher latency? I realize that the difference isn't very significant in the grand scheme of things, but it's always nice to shave a few ms off of initial page load times. -- Dave Warren http://www.hireahit.com/ http://ca.linkedin.com/in/davejwarren ___ Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe from this list bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users